The present invention relates to apparatus and methods for manufacturing integrated circuits and other electronic devices.
One of the basic problems in integrated circuit manufacturing is defects caused by the presence of particulates. For example, if photolithography with 0.8 micron minimum geometry is being performed to pattern a conductor layer, the presence of a 0.5 micron particle can narrow the patterned line enough to cause a defect which will prevent the circuit from operating (either immediately due to an open circuit, or eventually due to electromigration). For another example, if a 100 .ANG. particle of silicon adheres to the surface and is included in a 200 .ANG. nitride layer being grown, the dielectric will have greater chances of breaking down at that point, even assuming that no subsequent process step disturbs the silicon particle.
This problem is becoming more and more troublesome because of two trends in integrated circuit processing: First, as device dimensions become smaller and smaller, the size of a "killing defect" becomes smaller, so that it is necessary to avoid the presence of smaller and smaller particles. This makes the job of making sure that a clean room is really clean increasingly difficult. For example, a clean room which is Class 1 (i.e. has an atmosphere with less than one particle per cubic foot) for particles of one micron and larger may well be Class 1000 or worse if particle sizes down to 100 .ANG.ngstroms are counted.
Second, there is an increased desire to use large size integrated circuits. For example, integrated circuit sizes larger than 50,000 square mils are much more commonly used now than they were five years ago. This means that each fatal defect is likely to destroy a larger area of processed wafer than was previously true. Another way to think of this is that not only has the critical defect size decreased, but the critical defect density has also decreased.
Thus, particulates are not only an extremely important source of loss in integrated circuit manufacturing yields, but their importance will increase very rapidly in the coming years. Thus, it is an object of the present invention to provide generally applicable methods for fabricating integrated circuits which reduce the sensitivity of the process to particulate contamination.
One of the major sources of particulate contamination is humangenerated, including both the particles which are released by human bodies and the particles which are stirred up by equipment operators moving around inside a semiconductor processing facility (front end). To reduce the potential for particulate contamination from this major source, the general trend in the industry has been to make more use of automatic transfer operations. Using such operations, for example, a cassette of wafers can be placed into a machine, and then the machine automatically transfers the wafers, one by one, from the cassette through the machine (to effect the processing steps necessary) and back to the cassette, without manual assistance.
However, efforts in the area of automatic transfer operations have served to highlight the importance of a second source of particles, namely particles generated by the wafers and the transfer mechanisms during handling and transport operations. When the surface of the wafer jostles slightly against any other hard surface, some particulate (of silicon, silicon dioxide, or other materials) is likely to be released. The particulate density inside a conventional wafer carrier is typically quite high, due to this source of particulate. Moreover, many of the prior art mechanisms for wafer transport generate substantial quantities of particulate. The general problem is discussed in U.S. Pat. Nos. 4,439,243 and 4,439,244, which are incorporated by reference hereinto.
Some types of wafer processing are shown in U.S. Pat. Nos. 4,293,249 by Whelan issued on Oct. 6, 1981, 4,306,292 by Head issued on Dec. 15, 1981and 3,765,763 by Nygaard issued on Oct. 16, 1973, which are incorporated by reference hereinto.
The prior applications of common assignee discussed above addressed this facet of the problem by providing a vacuum wafer carrier in which particulate generation due to abrasion of the surface of the wafer during transport is reduced. The teachings of these prior applications enabled not only reduced generation of particulate in the carrier during transport and storage, but also reduced transport of particulate to the wafer's active face during transport and storage, by carrying the wafers face down under a high vacuum. This allowed the rapid settling of both ambient and transport generated particulate on other than the active wafer face.
The wafers can therefore be transported, loaded, unloaded and processed without ever seeing atmospheric or even low vacuum conditions. This is extremely useful, because, at pressures of less than about 10.sup.-5 Torr, there will not be enough Brownian motion to support particles of sizes larger than about 100 .ANG., and these particles will fall out of this low-pressure atmosphere relatively rapidly.
FIG. 2 shows the time required for particles of different sizes to fall one meter under atmospheric pressure. Note that, at a pressure of 10.sup.-5 Torr or less, even 100 .ANG. particles will fall one meter per second, and larger particles will fall faster. (Large particles will simply fall ballistically, at the acceleration of gravity.) Thus, an atmosphere with a pressure below 10.sup.-5 Torr means that particles one hundred angstroms or larger can only be transported ballistically, and are not likely to be transported onto the critical wafer surface by random air currents or Brownian drift.
The relevance of this curve to the various embodiments described in the present application is that the prior applications were the first known teachings of a way to process wafers so that the wafers are never exposed to airborne particulates, from the time they are loaded into the first vacuum process station (which might well be a scrubbing and pumpdown station) until the time when processing has been completed, except where the processing step itself requires higher pressures (e.g. for conventional photolithography stations or for wet processing steps). This means that the total possibilities for particulate collection on the wafers are vastly reduced.
The prior applications cited above also taught use of the vacuum wafer carrier design together with a load lock and vacuum wafer transport mechanism at more than one process module, to provide a complete low-particulate wafer transfer system. These vacuum load locks can usefully incorporate mechanisms for opening a vacuum wafer carrier after the load lock has been pumped down, for removing wafers from the carrier in whatever random-access order is desired, and for passing the wafers one by one through a port into an adjacent processing chamber. Moreover, the load lock mechanism can close and reseal the vacuum wafer carrier, so that the load lock itself can be brought up to atmospheric pressure and the vacuum wafer carrier removed, without ever breaking the vacuum in the vacuum wafer carrier. This process takes maximum advantage of the settling phenomena illustrated in FIG. 2 and described in more detail below. The wafer can then be moved in a virtually particulate free environment from the carrier to the load lock, into the process chamber and back through the load lock to the carrier for, potentially, an entire manufacturing sequence.
A process station (which may optionally contain one process module or more than one process module) has more than one load lock attached to it. This has several actual and potential advantages. First, processing can continue on wafers transferred in from one load lock while the other load lock is being reloaded, so that throughput is increased. Second, with some types of mechanical malfunction it will be possible to move at least the in-process wafers out of the central module area (into one of the load locks, or even into one of the process modules) to keep them from exposure to ambient if it is necessary to vent the process module to correct the malfunction. This means that even fairly severe faults may be recoverable. Third, if separate transfer arms are provided inside each of the load locks, this provides the further advantage that, if a mechanical problem occurs with one transfer apparatus inside its load lock, the process station can continue in production, using transfer through the other load lock, while maintenance is summoned to correct the mechanical malfunction.
The various process modules disclosed in the present application provide a tremendous improvement in the modularity of processing equipment. That is, a reactor can be changed to any one of a very wide variety of functions by a relatively simple replacement. It may be seen from the detailed descriptions below that most of the different functions available can be installed merely by making replacements in the wafer susceptor and related structures--i.e. in the top piece of the reactor, which bolts on--or in the feed structures. i.e. the structures directly below the wafer. Thus, the basic configuration of the vacuum chamber and wafer transfer interface is changed very little.
This capability confers tremendous advantages. First, the marginal capital cost of adding a new processing capability is greatly decreased. Second, the flexibility of manufacturing space is greatly increased, since machines can be reconfigured with relative ease to perform new functions. Third, the design development time for reactor structures is greatly decreased. Fourth, the time required to train personnel in use of a new reactor is also greatly decreased, since many key functions will be performed identically across a wide variety of reactors. Fifth, the cost of mistakes will be reduced, since operators will less frequently make mistakes due to unfamiliarity or confusion due to variety of equipment. Sixth, the carrying cost of an adequate spare parts inventory will be reduced. Seventh, the delay cost of repair and maintenance functions can be reduced, since many such functions can be performed off-line after an appropriate replacement module is swapped into the production reactor. Eighth, the presence of disused and obsolete machines in manufacturing space can be minimized, because a machine which had been configured to perform an unneeded function can be reconfigured.
The various classes of modules disclosed herein provide the advantage that the "footprint" required to emplace them is minimal. That is, if one or more process modules like those described is located in a clean room, only a minimum of clean room floor space (which is very expensive) will be required.
The capability for transferring wafers from one process chamber to another without breaking vacuum is enhanced by the modular compatibility of the below described embodiments. In particular, one of the advantages of modular processing units of the kind disclosed herein is that a single process station may advantageously contain several process modules like those described, so that wafers need not even go through the load lock to be transferred between two modules which are in a common station.
One way to think about the advantages of the various module designs discussed below might be to consider that they provide a super-capable reactor, i.e. has more adaptation capability than can ever be used for any single process. Viewed in this light, it may also be seen that their features are advantageous in sequential processing. That is, it has been recognized as desirable to perform more than one process in the same chamber without removing the wafer. The reactor designs disclosed herein are particularly advantageous in doing this, since the "excess" capability of the reactor design means that it is easier to configure it to perform two sequential steps.
Other and further advantages are set forth within and toward the end of the Description of the Preferred Embodiment.